Product Manager, AI Agents

AtomMatrix; San Francisco (CA); On-site; USD 150,000 - 210,000; Full time

Job summary

Atom Matrix Innovation LLC seeks aTechnical PM to own the AI agent platform roadmap, shaping instructions, tools, knowledge, guardrails, and cross-channel behavior. You will partner with AI and platform engineers to define scope, safety, and sequencing, turning high-stakes problems into tangible products for developers and operators.

You’ll work with customers to uncover wins and failures, write clear specs, and lead with data.

Qualifications

• 5+ years in product management, including technical or platform/API products.
• Fluency with how modern AI systems behave — you can have a real conversation with engineers about retrieval, tool calling, and evaluation.
• A track record of shipping products that developers or operators rely on, and of saying no well.
• Strong written communication — you make complex things clear on a page.
• Comfort with ambiguity and high stakes; you design for failure and human control, not just the happy path.

Responsibilities

• Own the vision, strategy, and roadmap for the AI agent platform, from tool definitions and knowledge to handoff and guardrails.
• Talk to customers constantly — understand where agents win, where they fail, and what would make teams trust them with more.
• Write the specs and make the trade-offs: what ships, what waits, and what we deliberately won't do.
• Partner with engineering on evaluation — define what "better" means and hold the bar with data.
• Work with design on how confidence, citations, approvals, and handoff appear to operators and end customers.
• Coordinate launches with docs, sales, and support so a capability lands well, not just ships.

Skills

Product management

APIs / API products

Customer interviews

Spec writing

Data-driven decisions

Cross-functional collaboration

Engineering collaboration

Communication

Job description

Agents are the heart of the platform and the hardest thing to get right. The line between "helpful" and "made something up," between "took the action" and "did the wrong thing," is a product problem as much as a model problem. We're looking for a PM who can own that line and the roadmap around it.

You’ll own the roadmap for AI agents — how customers define an agent’s instructions, tools, knowledge, and guardrails, how runs behave across channels, and when an agent hands off to a person.

This is a technical PM role. You’ll turn messy, high-stakes customer problems into a product that other builders can reason about, work shoulder to shoulder with AI and platform engineers, and make the hard calls about scope, safety, and sequencing. Success looks like agents that resolve more real cases while staying grounded and controllable — and customers who trust them enough to turn them on.

What you’ll do

• Own the vision, strategy, and roadmap for the AI agent platform, from tool definitions and knowledge to handoff and guardrails.
• Talk to customers constantly — understand where agents win, where they fail, and what would make teams trust them with more.
• Write the specs and make the trade-offs: what ships, what waits, and what we deliberately won't do.
• Partner with engineering on evaluation — define what "better" means and hold the bar with data.
• Work with design on how confidence, citations, approvals, and handoff appear to operators and end customers.
• Coordinate launches with docs, sales, and support so a capability lands well, not just ships.

What we’re looking for

• 5+ years in product management, including technical or platform/API products.
• Fluency with how modern AI systems behave — you can have a real conversation with engineers about retrieval, tool calling, and evaluation.
• A track record of shipping products that developers or operators rely on, and of saying no well.
• Strong written communication — you make complex things clear on a page.
• Comfort with ambiguity and high stakes; you design for failure and human control, not just the happy path.

Nice to have

• Background in CPaaS, customer support tooling, or conversational products.
• Hands-on technical experience (former engineer, or you still prototype).
• Experience with regulated or enterprise buyers.